Semantic process mining of enterprise transaction data

Abstract

Process mining technologies provide capabilities for discovering and describing multiple perspectives of the real business process flows in an organization. Enterprise Resource Planning (ERP) systems are commonly stated in research as promising areas for process mining. ERP systems are application packages that have received wide industrial adoption, and they contain extensive amounts of data related to business process performance. However, very little research work describes actual experience from applying process mining in such industrial environments.

In the work presented in this thesis, we have conducted studies on applying process mining techniques on real life ERP transaction data and we have explored technical opportunities targeting challenges introduced by the real world.

Specifically, this thesis answers the following four research questions:

RQ1. How can ontologies be applied to harmonize and interpret ERP transaction data?

RQ2. Can reliable business process traces be extracted from large-scale transaction logs in ERP systems?

RQ3. To what extent can semantic search techniques enrich process mining with explorative knowledge discovery?

RQ4. How can ontologies be used to lift process mining from the technical level to a conceptual business level?

The main contributions of this thesis are:

C1. Ontology driven harmonization of event log structures from ERP data.

C2. Ontology driven search for explorative investigations of process executions.

C3. Techniques for annotating unlabeled transaction sequences with business process definitions.

C4. Use of ontologies to manage perspectives of process mining models, define trace clusters and to extend the number of dimensions for data mining.

C5. Value of search and semantics in business process mining on ERP transaction data.